Current research projects

Carbon cycling in Arctic tundra

Arctic ecosystems contain significant carbon stocks, mostly frozen organic matter in soils. Predictions of global warming suggest the largest temperature rises may occur in high latitudes. A major research issue is whether this warming will result in a thawing of artic soils, and thus a stimulation of decomposition. If warmed soils start to break down and emit of CO₂ into the atmosphere, then global warming may be accelerated. Our research has focused on exploring the potential for global change to affect C cycling in tundra ecosystems. Our particular interest has been in scaling up understanding from intensive work at specific field sites, across the broader and varied arctic landscape.

Much of the underlying research has been undertaken at the Arctic Long-Term Ecological Research site at Toolik Lake, Alaska, and the Abisko Research Station of the Royal Swedish Academy of Sciences, and funded by the US National Science Foundation, Arctic System Science program.

Our research has demonstrated that vegetation in arctic ecosystems is organised according to a consistent pattern - the amount of leaf area is strongly related to the total investment in nitrogen in the plant canopy. Our findings in Alaska have now been replicated in Sweden (unpublished results), demonstrating the conservative nature of the relationship we have found. The relationship is important because it suggests a powerful homeostatic mechanism involved in determining the structure of arctic plant communities. We continue to investigate this mechanism in current research.

We have also shown that the variation in rates of C exchange across the arctic landscape during mid-summer can be largely explained by the variation in vascular plant cover. This result corroborates the model we use to make predictions of ecosystem C and energy exchange (the SPA model), but also suggests that the activity of mosses and lichens is probably concentrated in the early and late parts of the growing season.

We have used the SPA model to estimate rates of gross production (i.e.. photosynthesis) across a large arctic region. This procedure involved using data from satellites to characterise the land cover, weather station data to estimate climate across the region, and a series of alterations (or aggegrations) of the SPA model to allow it to be used in a regional mode. Our paper looks in detail at where the main problems arise in making regional predictions - identifying which assumptions or simplifications lead to the largest degree of uncertainty in the final estimate. We found that uncertainty in predicting land-cover from space-based instruments was a critical problem, as were difficulties in determining the variation in cloudiness across the landscape at sub-diurnal (i.e. hourly) time scales.

Understanding the timing of plant development (phenology) is critical for many ecosystem processes - for example, the development of leaf area determines at what point an ecosystem starts to assimilate carbon. We are interested in determining to what degree changes in climate will affect plant phenology. In arctic ecosystems, phenology is closely controlled by the timing of snow melt and soil thaw. Thus, to predict the impact of climate change, accurate models of snow and soil dynamics are required. In a paper currently in press with Global Change Biology, we have shown the the newest version of the SPA model is capable of predicting soil temperatures and snow pack over many years in arctic tundra. We also show that current models of phenology, which are generally very simple, predict very different predictions of bud break under climate change than does our soil thaw model.

We are currently writing papers exploring issues such as: the difference in response to fertilization of arctic ecosystems in Alaska versus Sweden; the importance of luxury uptake of nutrients in determining the impacts of fertilization on plant communities; and the effect of micro-topography on vegetation in arctic ecosystems.

Collaborators:

• Gus Shaver; The Ecosystems Centre, Marine Biological Laboratory
• Mark van Wijk, Wageningen University, Netherlands.

Tropical rain forests and global change

Tropical forests store 40 % of the carbon (C) residing in the world's terrestrial vegetation and annually process about six times as much C through photosynthesis as is released to the atmosphere by fossil fuel combustion. Relatively small changes in tropical forests could therefore have significant global consequences for the C cycle and the rate of climate change.

Global climate models have only recently begun to examine vegetation-climate feedbacks. Future climate may be highly dependent on the behaviour of tropical forests, if climate change shifts Amazonia from a C sink to a source by mid-century, as a result of drying and warming. However current climate models use relatively crude representations of vegetation, especially in tropical regions where field-based understanding is lacking.

Our research focuses on understanding the links between water and carbon cycling in tropical rain forest. We initially examined the controls on seasonal C exchange in a central Amazonian rain forest, indentifying a likely below-ground constraint on water flow, and this production, in the local dry season. See: Williams et al. (1998). Recently, global modelling studies have suggested that eastern Amazonia may be vulnerable to climate change (Cox et al, 2000). The modelling suggests that forest dieback may occur as rain fall declines, and that the conversion of forest to grassland may accelerate global warming by releasing extra CO₂ to the atmosphere. We tested this hypothesis, by artificially droughting a section of rain forest. We installed panels beneath the canopy on one hectare of forest to channel water away from the soil. The droughting began in early 2002 and has continued through 2010. We have monitored the behaviour of the trees as drought develops, and have used these results to improve our models of how forests respond to changing water availability.

As part of the Large-Scale Biosphere-Atmosphere Experiment in Amazonia, an international, Brazil-led project, we have made extensive surveys of vegetation and soils throughout the Tapajos National Forest, Para, in eastern Amazonia, with a view to investigating how heterogeneity affects uncertainties in scaling up ecological processes to the landscape and regional level. Survey results are described here.

Collaborators

• Patrick Meir, John Grace, University of Edinburgh
• Rosie Fisher, NCAR
• Yadvinder Malhi, University of Oxford

Upscaling Carbon exchange

Detailed knowledge of the processes underlying biogenic GHG emissions is vital for understanding, attributing and managing sinks and sources. Such understanding leads to improvements in flux estimates, allows inventories with greater sensitivity to management and climate, and advances the modelling of feedbacks between climate, land use and GHGs. Most information on biogenic fluxes of GHGs is obtained at small scales in detailed field studies. However, policy makers require information on GHG emissions at landscape, regional and national scales. Using the field-scale knowledge to infer estimates of emissions at large scales is, however, a major challenge. The scaling challenge arises because of inherent spatial and temporal variability in the underlying processes driving GHG exchanges. Models developed and parameterised at a fine scale cannot simply be translated to regional scales. Wherever model responses are nonlinear, the distribution of input variables needs to be known over the whole domain at the model resolution. Upscaling requires that sources and sinks are appropriately weighted, and that non-linear responses are correctly resolved

Collaborators

Bev LawOregon State University

Andy Fox, NCAR/NEON

Tim Hill, St Andrew's University

Plant traits and biomass production

Long-term predictions of ecosystem behaviour are particularly dependent upon the representation of feedbacks among water, carbon and nutrient cycles. We have focused particularly on understanding how requirements for water transport affect the activity of forest stands. The soil-plant-atmosphere model makes explicit the connection between liquid and vapour phase water transfer, tracking the dynamics of leaf water potential and its interactions with stomatal conductance. Because the model is based on underlying processes, the model can be used to diagnose complex datasets and predict potential future behaviour.

The original calibration and corroboration of the SPA model was against long-term eddy covariance data from the Harvard Forest, Massachusetts. The model has recently been upgraded to include more detailed representation of hydraulic processes below ground , with corroboration against a dataset from a temperate coniferous forest.

One issue of particular interest is the relationship between tree size and productivity. Many studies have reported an apparent decline in production as trees age, but the mechanisms are unclear. Working with Dr. Barbara Bond and Dr. Mike Ryan, we have modified the SPA model to simulate sap-flow in individual stems, and investigated alternative hypotheses of hydraulic limitation in a ponderosa pine stand.

Carbon Dynamics and Earth Observation - NCEO

I am a Principal Investigator with NCEO, a Natural Environment Research Council Centre of Excellence. My particular focus is on developing improved estimates of ecosystem C budgets, which is crucial given current uncertainties about the global carbon cycle and its future behaviour. My approach is to use the Kalman Filter to combine model and observations into an analysis with improved confidence bounds. The Kalman filter can also be used to provide best estimates of model parameters that are poorly defined, so it also serves as a research tool.

Recent outputs

1. We have developed a “dissaggregation” approach to make use of the excellent temporal sampling of coarse resolution satellite data, whilst also avoiding introducing bias. The disaggregation approach combines frequent coarse resolution observations with temporally sparse fine resolution measurements and conserves key ecosystem characteristics regardless of the observation resolution.

Hill, T. C., Quaife, T. and Williams, M. (2011) A data assimilation method for using low-resolution Earth Observation data in heterogeneous ecosystems. J. of Geophysical Research 116:D08117.

2. Carbon emissions from tropical land-use change are a major uncertainty in the global carbon cycle. We have presented a method for mapping vegetation carbon stocks and their changes over a 3-year period in a > 1000 km² region in central Mozambique at 0.06 ha resolution. L-band synthetic aperture radar imagery and an inventory of 96 plots are combined using regression and bootstrapping to generate biomass maps with known uncertainties. The resultant maps have sufficient accuracy to be capable of detecting changes in forest carbon stocks of as little as 12 MgC/ha/yr over 3 years with 95% confidence. This allows characterization of biomass loss from deforestation and forest degradation at a new level of detail.

Ryan, C. M., T. C. Hill, E. Woollen, C. Ghee, E. T. A. Mitchard, G. Cassells, J. Grace, I. H. Woodhouse, and M. Williams. 2012. Quantifying small-scale deforestation and forest degradation in African woodlands using radar imagery. Global Change Biology 18: 243-257.

3. We have determined to what extent can time series of CO₂ flux observations be used to calibrate models of carbon fluxes and improve their predictions. Through a synthetic study comparing two data assimilation approaches, we show that the length of an eddy flux CO₂ data set has more influence on the inversion process that either data gaps or observation noise. Model predictions were more effective with parameters estimated from long data time series. We also show that assumptions within the model have an effect of parameter estimation, causing ill posed problems in some cases. We suggest that modifying models to reproduce emergent relationships between carbon pools would help reduce this equifinality problem.

Hill, T.C., E. Ryan and M. Williams (2012) The use of CO₂ flux time series for parameter and carbon stock estimation in carbon cycle research. Global Change Biology 18: 179-193.

Collaborators

Casey Ryan

Mat Disney, Phil Lewis, UCL

Tim Hill, St Andrew's University